MASC: A Dataset for the Development and Classification of Mobile Applications Screens

Abstract

Abstract Mobile applications have become an integral part of our daily lives, offering a wide range of functionalities and services. Understanding the diversity of mobile application screens is crucial for optimizing user experience and delivering personalized content. This paper presents a novel dataset, called MASC (Mobile App Screens Classification) consisting of 7065 images, representing various types of mobile apps screens. MASC dataset is collected from the well-known Rico dataset. These screens were carefully manually classified into ten unique classes to capture the diverse nature of app interfaces. By employing advanced feature extraction techniques, we extracted key characteristics from each screenshot image of app screens related to visual elements, text, and keywords. Based on this dataset, this paper presents a proposed framework for applying machine learning algorithms to the classification of mobile apps screens. Using this framework, the paper also presents a comprehensive study of the classification of mobile apps screens using machine learning algorithms. Several classification algorithms including XGBoost, Gradient Boosting, Random Forest, SVM, Logistic Regression, and others were trained and evaluated on MASC. Results showed high accuracy rates above 93% for top models like Gradient Boosting, indicating that machine learning provides an effective approach to mobile app screen classification. This study contributes to the field of mobile application analysis and user interface understanding. In addition, the proposed mobile app screens classification framework is a promising development that can enhance the accuracy and efficiency of mobile app screens classification.